Ruby: flip between require and require_relative in minitests

TBH as a gem maintainer I have never needed to run gem tests against some specific release. But it seems the following approach should work in your case:

• use require "<my-gem>" in tests
• run tests with bundler - bundle exec rake ...
• have 2 Gemfile files to run tests with local source code and with installed locally gem

Developing a gem you usually include dependencies declared in a gemspec file into a Gemfile file with the gemspec declaration:

# Gemfile

gemspec

# development dependencies
# ...

It adds into a Gemfile.lock file declaration to use source code of the developed gem (my-gem) from source code located in the current directory. For instance for a gem I am maintaining (dynamoid) it looks like the following:

PATH
  remote: .
  specs:
    dynamoid (3.11.0)
      activemodel (>= 4)
      aws-sdk-dynamodb (~> 1.0)
      concurrent-ruby (>= 1.0)

More details you may find here - https://bundler.io/guides/rubygems.html.

In the second Gemfile to run tests agains a released version of your gem you may require explicitly my-gem with a specific version instead of using gemspec method:

# Gemfile

gem "<my-gem>", "1.0"

# development dependencies
# ...

As an option you can use a single Gemfile and check some env variable to either use gemspec method or declare explicitly a dependency on my-gem:

# Gemfile

if ENV['MY_GEM_VERSION']
  gem 'my-gem', ENV['MY_GEM_VERSION']
else
  gemspec
end

# development dependencies

Usually I use the following dirty hack to run some gem's tests locally:

1. install dependencies with bundle install command
2. use ruby -I ./lib -I ./test rake test to run tests agains the source code
3. use rake test to run tests agains an installed gem version